Agent Matching
Human-readable runbook for a recurring automation loop — an AI agent runs one pass on a schedule, discovering its own work each time. Same shape as a Process Memo (summary, purpose, prerequisites, ordered steps, conventions) scoped to one repeatable pass, with idempotency rules, a cheap empty pass, and a declared runtime posture in the Conventions. Use for memo-drafting loops, monitoring loops, and other scheduled agent automations. Not the executable body a MythOS Loop runs — that is a loop definition. Deciding question vs the Process Memo Template: does an agent discover its own work on a timer (loop), or does a human hand it work (process)?
No tags.